Machine learning models using chest CT and clinical data predict lung cancer subtype for early brain metastasis therapy

ML models integrating clinical data and chest CT features predicted lung cancer subtype (small cell, EGFR-mutant, EGFR-wild-type) with AUC up to 0.907, to guide early brain metastasis management.

• Retrospective, single-center study with a training cohort of 182 patients (no brain metastases) and a testing cohort of 123 patients presenting with synchronous brain metastases.
• Random forest achieved AUC 0.907 for EGFR-mutant NSCLC; logistic regression AUCs were 0.887 (EGFR+), 0.811 (SCLC), and 0.801 (EGFR-).
• Adding imaging features (fibrosis, emphysema, miliary pattern, cavitation, pleural attachment, vessel encasement) to clinical variables improved prediction for EGFR+ and SCLC.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.